The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess calls to manage its environment and perform audio conversion. In scripts/generate_audio.py, ffmpeg is called to convert WAV files to MP3 using hardcoded parameters and controlled file paths. In scripts/setup_environment.py and scripts/run.py, subprocess is used to manage the Python virtual environment and install packages. These operations are standard for the skill's architecture and do not exhibit command injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill downloads Python dependencies from PyPI. It employs a requirements.lock file with SHA-256 hashes for all packages (e.g., google-genai, pydantic, cryptography), ensuring that only verified, untampered code is installed in the local environment.
[DATA_EXFILTRATION]: The skill communicates with official Google API endpoints via the google-genai SDK. This is the intended behavior for text-to-speech generation. It correctly handles the API key via environment variables and does not transmit sensitive local files to untrusted domains.
[PROMPT_INJECTION]: The skill processes user-provided blog content for narration. While this creates a surface for indirect prompt injection (narrating malicious text), the impact is confined to the generated audio output and does not grant the attacker unauthorized capabilities.

blog-audio